MLLP-UPV and RWTH Aachen Spanish ASR Systems for the IberSpeech-RTVE 2018 Speech-to-Text Transcription Challenge

Jorge, Javier; Martinez-Villaronga, Adria A.; Golik, Pavel; Giménez, Adrià; Silvestre-Cerdà, Joan Albert; Doetsch, Patrick; Císcar, Vicent Andreu; Ney, Hermann; Juan, Alfons; Sanchís, Alberto

doi:10.21437/iberspeech.2018-54

Cited by 7 publications

(21 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To carry out our experiments, we used the development and test sets from the RTVE2018 database. More precisely, we devoted our internal dev1-dev set [4] for development purposes, whilst dev2 and test-2018 were dedicated to test ASR performance. Finally, test-2020 was the blind test used by the organisation to rank the participant systems.…”

Section: Experiments and Resultsmentioning

confidence: 99%

“…At this point, we defined our participant off-line hybrid ASR system identified as c3-offline (contrastive system no. 3), consisting of a fast pre-recognition + Voice Activity Detection (VAD) step to detect speech/no-speech segments as in [4], followed by a real-time one-pass decoding with our BLSTM-HMM AM, using a FSN normalization scheme and a linear combination of the three types of LMs: n-gram, LSTM and Transformer. This system scored 12.3 and 17.1 WER points on test-2018 and test-2020, respectively.…”

Section: Experiments and Resultsmentioning

confidence: 99%

“…The MLLP-VRAIN participated in the 2018 edition of the challenge [4] in a joint collaboration with the Human Language Technology and Pattern Recognition (HLTPR) research group from the RWTH Aachen University. The evaluation was carried out on the RTVE2018 database [5], that includes 575 hours of audio from 15 different TV shows broadcasted between 2015 and 2018.…”

Section: Challenge Description and Databasesmentioning

confidence: 99%

“…Our acoustic models (AM) were trained using 205 filtered speech hours from the train set (187h) and our internal dev1train set (18h), as in [4], plus about 3.7K hours of other resources crawled from the Internet. Table 1 summarises all training datasets along with their total duration (in hours).…”

Section: Acoustic Modellingmentioning

confidence: 99%

See 3 more Smart Citations

MLLP-VRAIN Spanish ASR Systems for the Albayzin-RTVE 2020 Speech-To-Text Challenge

Jorge¹,

Giménez²,

Baquero-Arnal³

et al. 2021

IberSPEECH 2021

View full text Add to dashboard Cite

This paper describes the automatic speech recognition (ASR) systems built by the MLLP-VRAIN research group of Universitat Politècnica de València for the Albayzin-RTVE 2020 Speech-to-Text Challenge.The primary system (p-streaming 1500ms nlt) was a hybrid BLSTM-HMM ASR system using streaming one-pass decoding with a context window of 1.5 seconds and a linear combination of an n-gram, a LSTM, and a Transformer language model (LM). The acoustic model was trained on nearly 4,000 hours of speech data from different sources, using the MLLP's transLectures-UPV toolkit (TLK) and TensorFlow; whilst LMs were trained using SRILM (n-gram), CUED-RNNLM (LSTM), and Fairseq (Transformer), with up to 102G tokens. This system achieved 11.6% and 16.0% WER on the test-2018 and test-2020 sets, respectively. As it is streaming-enabled, it could be put into production environments for automatic captioning of live media streams, with a theoretical delay of 1.5 seconds.Along with the primary system, we also submitted three contrastive systems. From these, we highlight the system c2streaming 600ms t that, following the same configuration of the primary one, but using a smaller context window of 0.6 seconds and a Transformer LM, scored 12.3% and 16.9% WER points respectively on the same test sets, with a measured empirical latency of 0.81±0.09 seconds (mean±stdev). This is, we obtained state-of-the-art latencies for high-quality automatic live captioning with a small WER degradation of 6% relative.

show abstract

Section: Experiments and Resultsmentioning

confidence: 99%

Section: Experiments and Resultsmentioning

confidence: 99%

Section: Challenge Description and Databasesmentioning

confidence: 99%

Section: Acoustic Modellingmentioning

confidence: 99%

See 2 more Smart Citations

MLLP-VRAIN Spanish ASR Systems for the Albayzin-RTVE 2020 Speech-To-Text Challenge

Jorge¹,

Giménez²,

Baquero-Arnal³

et al. 2021

IberSPEECH 2021

View full text Add to dashboard Cite

show abstract

“…The language model was a five-gram. • G7-MLLP-RWTH [21]. MLLP, Machine Learning and Language Processing, Universidad Politécnica de Valencia, Spain and Human Language Technology and Pattern Recognition, RWTH Aachen University, Germany.…”

Section: Open-set Condition Systemsmentioning

confidence: 99%

Albayzin 2018 Evaluation: The IberSpeech-RTVE Challenge on Speech Technologies for Spanish Broadcast Media

et al. 2019

View full text Add to dashboard Cite

The IberSpeech-RTVE Challenge presented at IberSpeech 2018 is a new Albayzin evaluation series supported by the Spanish Thematic Network on Speech Technologies (Red Temática en Tecnologías del Habla (RTTH)). That series was focused on speech-to-text transcription, speaker diarization, and multimodal diarization of television programs. For this purpose, the Corporacion Radio Television Española (RTVE), the main public service broadcaster in Spain, and the RTVE Chair at the University of Zaragoza made more than 500 h of broadcast content and subtitles available for scientists. The dataset included about 20 programs of different kinds and topics produced and broadcast by RTVE between 2015 and 2018. The programs presented different challenges from the point of view of speech technologies such as: the diversity of Spanish accents, overlapping speech, spontaneous speech, acoustic variability, background noise, or specific vocabulary. This paper describes the database and the evaluation process and summarizes the results obtained. in Spanish [8][9][10][11], and more recently, the Multi-Genre Broadcast (MGB) Challenge with data in English and Arabic 2 [12][13][14]. In other areas apart from broadcast speech, several evaluation campaigns have been proposed such as the ones organized in the scope of the Zero Resource Speech Challenge [15,16], the TC-STAR evaluation on recordings of the European Parliament's sessions in English and Spanish [5], or the MediaEval evaluation of multimodal search and hyperlinking [17].As a way to measure the performance of different techniques and approaches, in this 2018 edition, the IberSpeech-RTVE Challenge Evaluation campaign was proposed in three different conditions: speech-to-text transcription (STT), speaker diarization (SD), and multimodal diarization (MD). Twenty-two teams registered to the challenge, and eighteen submitted systems in at least one of the three proposed tasks. In this paper, we describe the challenge and the data provided by the organization to the participants. We also provide a description of the systems presented to the evaluation, their results, and a set of conclusions that can be drawn from this evaluation campaign.This paper is organized as follows. In Section 2, the RTVE2018 database is presented. Section 3 describes the three evaluation tasks, speech-to-text transcription, speaker diarization, and multimodal diarization. Section 4 provides a brief description of the main features of the submitted systems. Section 5 presents results, and Section 6 gives conclusions.

show abstract

Live Streaming Speech Recognition Using Deep Bidirectional LSTM Acoustic Models and Interpolated Language Models

Jorge

Giménez

Silvestre-Cerdà

et al. 2022

IEEE/ACM Trans. Audio Speech Lang. Process.

Self Cite

View full text Add to dashboard Cite

Although Long-Short Term Memory (LSTM) networks and deep Transformers are now extensively used in offline ASR, it is unclear how best offline systems can be adapted to work with them under the streaming setup. After gaining considerable experience on this regard in recent years, in this paper we show how an optimized, low-latency streaming decoder can be built in which bidirectional LSTM acoustic models, together with general interpolated language models, can be nicely integrated with minimal perfomance degradation. In brief, our streaming decoder consists of a one-pass, real-time search engine relying on a limited-duration window sliding over time and a number of ad hoc acoustic and language model pruning techniques. Extensive empirical assessment is provided on truly streaming tasks derived from the well-known LibriSpeech and TED talks datasets, as well as from TV shows on a main Spanish broadcasting station.

show abstract

MLLP-UPV and RWTH Aachen Spanish ASR Systems for the IberSpeech-RTVE 2018 Speech-to-Text Transcription Challenge

Cited by 7 publications

References 13 publications

MLLP-VRAIN Spanish ASR Systems for the Albayzin-RTVE 2020 Speech-To-Text Challenge

MLLP-VRAIN Spanish ASR Systems for the Albayzin-RTVE 2020 Speech-To-Text Challenge

Albayzin 2018 Evaluation: The IberSpeech-RTVE Challenge on Speech Technologies for Spanish Broadcast Media

Live Streaming Speech Recognition Using Deep Bidirectional LSTM Acoustic Models and Interpolated Language Models

Contact Info

Product

Resources

About